Hoi,
The point is very much that where we did not have these images at all, we are now blessed with having them, When you read the article, you will read that the project was about going back to the scans and extract the images from it. Those scans were used before to get only the text out.
So in a round about way, they did have the meta data about the images and extracted them. They have the text and now they now have the images as well. What is left is stitching the images and the texts back together based on the existing meta data.. In essence everything to make this happen is available.
Thanks,
GerardM